«Department of Economics, Mathematics and Statistics BWPEF 1408 Incumbency Advantage at Municipal Elections in Italy: A Quasi-Experimental Approach ...»
Birkbeck Working Papers in Economics & Finance
Department of Economics, Mathematics and Statistics
Incumbency Advantage at Municipal
Elections in Italy: A Quasi-Experimental
Marco Alberto De Benedetto
Birkbeck, University of London
▪ Birkbeck, University of London ▪ Malet Street ▪ London ▪ WC1E 7HX ▪
Incumbency Advantage at Municipal Elections in Italy: A Quasi-
Marco Alberto De Benedetto
Department of Economics, Mathematics and Statistics, Birkbeck College Abstract We analyze the incumbency effect on a candidate’s electoral prospects using a large data set on Italian municipal elections held from 1993 to 2011. We apply a non-parametric Sharp Regression Discontinuity Design that compares candidates who barely win an election to those who barely lose, exploiting the fact that incumbency status changes discontinuously at the threshold of margin of victory of zero. We find that incumbents are more likely to win the competition compared to their challengers at the Italian municipal elections. The results are robust to different specifications and estimation strategies with excellent balance in observable characteristics. Also, the effect of interest seems to be larger in magnitude for municipalities located in the North of Italy compared to southern municipalities.
JEL Classification: D72, D78; J45 Keywords: Incumbency Status; Political Participation; Sharp RDD.
1. Introduction A growing body of literature emphasizes the crucial role played by political institutions in the process of economic development and in particular, how the distribution of political power2 improves economic performance and determines the allocation of resources, even in a democracy (Acemoglu, 2002). According to the fiscal common theory (Tullock, 1959; Buchanan and Tullock, 1962), politicians once elected, in order to obtain a greater electoral support, try to use their influence to redirect resources to particular groups of constituents to the detriment of the general community. The fact that holding political power makes such a difference is the reason that democratic governments are founded on the principle that voters should ultimately decide which representatives are chosen to wield power3 (Pande, 2003).
1 I would like to thank Walter Beckert and Maria De Paola for useful comments.
2 The political power is defined as an authority held by a group within a society that allows for the administration of public resources and implements policies for the society. Distribution of power is in balance when each decision is made by the group of individuals (politicians and citizens) affected by the consequences of the decision per se. Acemoglu (2002) shows that inefficient institutions and policies are chosen because they serve the interests of politicians or social groups that hold political power at the expense of the rest. Related to this argument, Acemoglu and Robinson (2000b, 2002) explain why rulers who fear replacement may pursue the wrong policies for the society. In that paper, rulers who fear replacement are more likely to resist the introduction of superior technologies or institutions when these changes may erode their incumbency advantage and their potential future political power.
3 Further, since long tenure in public office leads to abuse of power, legislators in many democracies are subject to a term limit.
1 The major risk in a democracy is that elected officials will become entrenched or that running for office may simply become too expensive for fresh-candidates. By the nature of the democratic system, being incumbent is intrinsically advantageous since he/she is given access to resources and decision processes that non-incumbent challengers do not have. If elected officials are able to use their political influence to remain in power, voters will have a limited influence on their policy decisions (Linden, 2004), especially where incentives to engage in rent extraction usually run high (Titiunik, 2011). Moreover, stronger incumbents also raise the cost of entering politics and reduce the degree of political competition because new challengers might not have enough resources to overcome the advantage of incumbency and as a consequence, voters might be less inclined to participate at the polls.
For this reason, a large literature has investigated the effect of the incumbency status on the probability that an incumbent candidate wins the electoral competition both at the state (Garand, 1991; King, 1990; Cox and Morgenstern, 1993) and federal level (Erikson, 1971; Alford and Hibbing, 1981; Alford and Brady, 1988; Gelman and King, 1990) in U.S. House elections. Results generally show a personal incumbency advantage, defined as the votes gained by a candidate once he/she becomes an incumbent from constituency service, name recognition, and the like, in terms of winning the electoral competition.
Moreover, some authors have focused on the incumbency effect at national and state elections in developing democracies, since the likelihood to observe both entrenched politicians controlling the political process as well as rampant corruption is higher (Linden, 2004). However, the evidence from some developing countries, such as India (Linden, 2004; Uppal, 2009), Latin America and Caribbean countries (Molina, 2001)4, suggests that there is a disadvantage to incumbents. The only exception is Miguel and Zaidi’s (2003) investigation of national elections in Ghana in which they find no significant incumbency effect at the parliamentary seat level5.
Furthermore, related to this argument, a second strand of literature, following Lee (2008)’s work, has concentrated on the partisan incumbency effect, i.e. the electoral benefit a candidate receives purely because his/her party is the incumbent party, regardless of whether he/she previously served (Fowler and Hall, 2012), highlighting again mixed results. In fact, Ferreira and Gyourko (2009) and Hainmueller and Kern (2008) find that the partisan incumbency status positively affects the probability of re-election and the likelihood of winning the competition in the US and in German districts at federal elections respectively. Conversely, Titiunik (2011), by using the same methodology 4 Molina (2001) argues that incumbent turnover is much higher in Latin American and Caribbean countries than many industrialized countries owing to endemic popular discontent over persistent deprivation. Conversely, for Indian national elections the incumbency disadvantage, especially after 1991, is essentially due to a change in the political structure that leads to a system in which as politicians gained more experience and influence they become more likely to pursue activities that are not in the best interest of voters (Linden, 2004).
5 Miguel and Zaidi (2003) justify their results saying that the lack of a meaningful incumbency advantage is consistent with a political system where the ruling party does not have adequate mechanisms at its disposal to accurately target funds down to the level of parliamentary seats. However, they acknowledge some important limitations of their data set, including the small sample size which leads to statistically imprecise estimates.
2 as Lee (2008), analyses the incumbency effect for three different political parties at Brazilian municipal elections held in 2000 and finds a negative effect of the partisan incumbency both on the incumbent parties’ votes share as well as on their probability of winning the competition.
In this paper we provide new evidence of the incumbency effect on a candidate’s electoral prospects, using a large data set on Italian municipal elections held from 1993 to 2011. Although many papers have focused on state and federal elections, maybe because politicians and voters attach a greater degree of importance and weight to national than to local elections arguing that local elections rarely make much difference in political life (Koryakov and Sisk, 2003) we study the personal incumbency advantage at municipal elections since they have certain distinct characteristics as compared with national elections which give them considerable significance in political life. First, local elections are important for their role in a broader national democracy, since their results are indicative of broader political trends and provide important information about the preferences, concerns and attitudes of the electorate. Second, issues in local elections are those that directly affect the daily lives of citizens: the nature of the competition between candidates and the issues that arise can be important indicators of what voters care deeply about and want the local authorities to tackle.
As far as the methodology implemented to recover the causal incumbency effect is concerned, the main difficulty in empirically estimating incumbency advantage/disadvantage is omitted variable bias, since some candidate’s characteristics such as charisma, charm and intelligence are typically unobservable and unquantifiable (Levitt, 1994). If higher quality candidates attract more votes, electoral selection will lead to incumbents and challengers possessing different characteristics. Failure to control for these differences may lead to biased estimates of incumbency advantage (Gelman and King, 1990). In order to overcome the problem of selection bias and omitted variable bias, we implement a regression discontinuity design (RDD). We focus on very close elections which are decided by a narrow margin of victory. The bare winners and bare losers of these elections are assumed to be comparable in all their baseline characteristics (we discuss this assumption in SubSection 3.1). This implies that bare losers provide a valid counterfactual for bare winners with regard to subsequent electoral outcomes. By comparing these outcomes we identify the causal effect of the incumbency status.
We believe that the contribution of our paper is twofold. Firstly, the literature investigating the incumbency advantage/disadvantage for European countries is scant, and in particular, no author so far has analysed whether incumbent mayors have or do not have an advantage in winning the competition compared to non-incumbent candidates for Italian municipal elections. In fact, decisions made at municipal level in Italy have a great impact on citizens’ daily lives, since these decisions often concern relevant services, such as the management of public utilities, the provision of public housing etc. For this reason, citizens are usually interested in the composition of the Municipal bodies and in the performance of the mayor, especially when he/she has already performed the same charge in the past. In particular, our Sharp Regression Discontinuity estimates show that the personal 3 incumbency effect, after controlling both for candidates and municipalities’ characteristics as well as for partisan swing and partisan incumbency, is about 34.3 percentage points, implying that incumbents (bare winners) are more likely to win the competition compared to their challengers (bare losers). Similar results are found when we use as dependent variable the vote share at time t (the personal incumbency effect is roughly 17 percentage points)6, after controlling for municipalities and candidates’ characteristics. Moreover, our findings are in line with those found in the literature (see for instance Alford and Brady, 1988; Gelman and King, 1990), although the methodology used in the previous papers does not take into account potential omitted variable bias. However, the effect we find is larger in terms of magnitude, maybe because we focus our analysis on local elections where the incumbency status explains most of the variation in the probability of winning the electoral competition compared to federal or national elections. To the best of our knowledge, the only authors using the Regression Discontinuity Design to find the causal effect of the personal incumbency status on the probability of winning the electoral competition are Uppal (2009) who finds an incumbency disadvantage of 22 percentage points over the probability of winning at time t after 1991 at Indian state elections, and Trounstine (2011) who highlights an incumbency advantage of 32 percentage points at city council elections between 1915 and 1985 in four U.S. cities.7 Secondly, we study the personal incumbency effect by taking into account the differences in the economic and social conditions of the two main geographical areas (South and North) in Italy. On the one hand, as the southern part of the country is poorer and endowed with a low level of social capital, the positive impact of the incumbency status on both the votes share as well as on the probability of winning the election may be related to the clientelistic relationships established by the incumbent candidates, which ensure political support in exchange of benefits (exchange votes). Areas endowed with low social capital are characterized by relationships that often involve requests for jobs and patronage, and citizens living in these areas may be more inclined to cast their vote in relation to exchange agreements (Knack, 2002). On the other hand, in areas characterized by general discontent as the South of Italy, incumbents may have some difficulties in satisfying the majority of voters and as a consequence, the probability of winning the electoral competition for incumbent candidates is expected to be lower than that of incumbents holding power in the North. Our findings are in line with this second explanation since bare winners are 42.5 percentage points more likely to win the competition compared to bare losers in the North, whereas for southern municipalities we find an incumbent advantage of 26.7 percentage points.
Finally, our results are robust to different specifications of our main equation. In particular, our findings are similar when we consider only observations in narrow neighborhoods around the discontinuity point (5 and 2 percent above and below the threshold of margin of victory of zero 6 Results not displayed, but available upon request.
7 As recognised by the author, these four cities (Austin, Dallas, San Antonio and San Jose) are not a representative sample of U.S. cities. Therefore, the conclusion of this paper may be limited by the sample used. Moreover, she does not have enough observations to separately analyze mayoral elections.
4 respectively), and when we choose different polynomials of the forcing variable (Local Linear Regression) along with the interaction terms between polynomials of the electoral margin (until the third-order) and the treatment.